The Structure of First-Species Canon in Modal - Tonal An Atonal MusicsMorris

The Structure of First-Species Canon in Modal,
Tonal and Atonal Musics
Robert D. Morris
It has been frequently and appropriately noted that serial

composition has affinity with pre-tonal polyphony. Certainly
both practices take the combinations of lynes of pitch-classes
as their structural frame, and a distinct lack of harmonic tele-
ology heightens the connection. But since fundamental re-
search into modal counterpoint has been rare in the twentieth
century, notwithstanding Serge Taneiev's brilliant counterpoint
treatise of 1909 and a few articles by American theorists,1
there have been few studies that relate the two practices in any
deep structural way. This paper attempts to bridge this
pre/post-tonal gap by studying first-species canons, primarily
at the delay of one note, using a single methodology reinter-
preted as necessary to accommodate modal, tonal and post-
tonal sounds and concepts.2
Admittedly, the limitation to studying first-species canon
restricts the scope of this paper. For, aside from simple four-
part chorales, note-against-note textures are hardly a norm in
music literature. Nonetheless, the study of first-species
polyphony addresses both primary and advanced concerns in
composition and analysis. From the Renaissance to the pre-
sent, the note-against-note contrapuntal ideal has been asserted
as fundamental to the study of both modal and tonal counter-
point and, in the 20th century, it has provided a methodolog-
ical base for doing tonal (Schenkerian) analysis. Canon, on
the other hand, is an advanced topic in modal counterpoint,
taken up only after all five species have been mastered. This
contrasts with post-tonal serial music where canon is primary,
since all aspects of serial polyphony may be considered
*See Serge Taneiev, Convertible Counterpoint (Moscow, 1909).

2Robert Gauldin has independently proposed a set of paradigms for compos-
ing and analyzing modal canons some of which are similar to our principles
for generating canons. See his "The Composition of Late Renaissance
Stretto Canons," Theory and Practice, forthcoming.
This content downloaded from 128.151.124.135 on Sat, 16 Mar 2019 00:11:38 UTC
All use subject to https://about.jstor.org/terms
6
1
«
CO
!
I
CO
C0
t
cd
X
to
>
I
I
<
cd
X
PQ
36 Integral
abstract canons at various intervals, inversions, retrogrades, and

delays.
But despite the importance of first-species texture and
canon in music pedagogy and analysis, it is difficult to find
musical examples that combine both canon and one-to-one
polyphony. First species canons find their only uses in the
structural foundation for close canons in the strettos of fugues
or in vorimitation of Baroque chorale preludes. Example la
shows that a first-species canon at the third below and delay of
one beat underlies a stretto in a Bach fugue. In serial music,
rotational arrays of pitch-classes, when interpreted homophon-
ically, form another example of first-species canon.3 See Ex-
ample lb for a rotational array underlying the final measures
of Stravinsky's Variations. The leading and following pcs
(which do not occur in the work) show how such an array is
identical to the middle segment of a pc canon. But despite
their rarity in the literature, as we shall see, first species canons
provide insight into the art of canon in older music as well as
offer today's composers some interesting compositional re-
sources.
Example 2. Canon at the fourth below and at a delay

of 1 beat:
3 See Robert Morris, "Generalizing Rotational Arrays," Journal of Music

Theory, 32/\ (1990).
Integral 37
Consider Example 2. It contains a note-agairist-note

canon a fourth below at the delay of 1 beat; I use the tradi-
tional terms dux for the leading voice, and comes for the fol-
lowing voice. The only vertical intervals are thirds and sixths,
and the only melodic intervals are the ascending third and the
descending second. The serial labels P and T.4P are used to
denote the relation between the voices.
Note that, as usual, we use integers to denote the tonal in-
tervals. For melodic ordered intervals: 1 is a note repetition, -4
a descending fourth, +3 (or 3) an ascending third, and so
forth. Distinctions between major and minor or augmented
and diminished intervals are not needed at this point. In two-
voice counterpoint vertical intervals are ordered, from the
"lower" voice to the "upper" voice. Therefore, vertical intervals
(that is, intervals produced by two simultaneous notes, each
from a different voice) are negative when voices are crossed.
So: 1 is a unison, 2 is a second, 3 is a third, and so on when
voices are uncrossed; when voices are crossed a vertical third is
given by -3, a vertical fifth by -5, et cetera. As is well-known,
the traditional interval names produce problems when adding
intervals. For instance, a third and a sixth "sum" to an octave,
but 3 + 6 = 9.4 Despite this drawback, here we use the tradi-
tional names to promote comprehension.
As Example 3 shows, we can describe the features of any
two- voice canon as an ordered quintuple:
(P,N,D,I,J)
where P is the name of the canon tune or subject, N is the

pitch interval of entry between the dux and comes, D is the
time delay, I is the set of vertical intervals between the voices,
and J is the set of linear or melodic (ordered) intervals in the
tune. As shown in Example 3, (P, -4, 1, {3, 6}, {+3, -2}) de-
scribes the canon in Example 2.
4More problems arise when negative values are also employed. The follow-
ing is the rule for addition of intervals when conventional interval names
are used: Let x and y be intervals, with lxl>=lyl. If x = -y, then x+y is defined
to be 1. Otherwise, x+y is defined as x+y-1 if y>0, and x+y+1 if y<-l.
38 Integral
Example 3: Canon descriptors.
General canon descriptor: (P, N, D, I, J)
P = canon subject
N = interval between dux and comes
D = time delay (in beats)

I = set of vertical intervals in canon
J = set of linear (melodic) intervals in P
Descriptor for canon in Example 2:

(P,-4,l,{3,6},{-2,+3})
In modal and tonal music, the vertical intervals allowable

for inclusion in set I are the same whatever the pitch interval
or delay; these are the consonant intervals {1, ±3, ±5, ±6, ±8,
±10, ±12}. Similarly, the linear intervals suitable for this music
(for inclusion in J) are drawn from the set {1, ±2, ±3, ±4, ±5,
±6, ±8}.5 Intervals larger than 3 are context dependent; for
instance, the melodic sequence +4, +4, is not allowed.
As any novice soon finds out, the limitation to these in-
tervals imposes severe constraints on the production of correct
counterpoint, especially in the composition of canons. These
constraints, which we will eventually display via a graph, are
dependent on N, D, I and J. Then, if a subject P can be derived
by following a connected path on such a graph, P will gener-
ate a correct canon at N and D.
Consider Example 4a, the simplest canon fragment of two
notes. The dux presents the notes a then b. This linear interval
from a to b is called j. The comes starts its note c with b of the
dux and then finishes with d; it also produces the interval j.
Now, D is 1 and N is the interval from a to c (or b to d). The
vertical interval i is the interval from b to c. From all of this, N
equals the sum of intervals j and i. Since the legal values of i
^ Note that in modal counterpoint, descending leaps of the sixth are discour-
aged. So +6 would be substituted for ±6 if we were to confine our study to
modal polyphony.
Integral 39
are specified in tonal music, they control the values of j for a

given N. Example 4b provides some examples of values for i
and j where N is +5 and D = 1 that generate valid canon frag-
ments, those that obey the rules of modal or tonal counter-
point. Note that some linear intervals (j) are ruled out by the
choice of N. With N = +5, for instance, j cannot be +2, for i
would have to be +4, a dissonance. Note also that i will be
negative during crossed voices, when note b is higher than
note c.
Example 4a. A minimal canon fragment.
Example 4b. Some values for j and i that generate

valid canon fragments at N = +5.
To generate canons longer than that of Example 4b, we

simply concatenate two linear intervals that satisfy the con-
straints of i, j, and N. Example 5a is a chart of all the possible
successions of two intervals for N = +5. The melodic intervals,
j, are -8, -6, -4, -2, 1,3 and 5. These are given by the row and
column heads of the chart; each cell of the chart, called a case,
contains a canon fragment generated by the interval succes-
sion of its row interval, then column interval. (The case of the
I
C/5
>">
•?
I
(L)
•5
X
I
PQ
p
$3 oo
2"
O c <D
<u ^-> cd >
8 ^ST
c
1 z 53 a
i I d
"S "I S
w) o in
O cd
<u uh is
C 3 C
^3 Q O
Vh O °
§° °o §&,
1
II II II
u o &
42 Integral
linear interval succession -4, 3 that generates a three-note

canon at the fifth above is therefore found in the cell in the -4
row and the 3 column of the chart.)
Example 5b. Graph for canon at fifth above.
-6 -4
Q . \ \/ \ "^ f^ /
v^^ . \ /\ X/1 f^ ^*>^*
Nv \ \l /^^^^ 1 ^^\
adjacency table
int #to #from ^ T
-8 2 2 \J^
-6 2 1
-4 3 3
_2 4 4 * = repeatable arrow
15 6
3 6 6
5 0 0
Of course, not all cas

canons - canons that me
16th-century or 18th-ce
-4 generates a highly un
parallel octaves. We theref
criteria given in the leg
polyphony that are us
counterpoint and, of co
Integral 43
fifths.6 After pruning, only valid cases are left; these are
written in half-note values while the invalid cases are written in
quarter-notes. From the chart we learn that the melodic
interval -6 can form a valid canon fragment if and only if it is
followed by the interval 3. The same interval may be only
preceded by the intervals 1 or 3. Such interval succession can
be illustrated by the use of a graph as in Example 5b. The
graph's nodes are the melodic intervals from which the canon
subject is constructed. The graph shows every legal case from
interval i\ to i2 on the chart, by an arrow connecting ii to 12
on the graph. In other words, the arrows show which intervals
may be concatenated to form a subject that automatically
obeys the rules of counterpoint inscribed in the chart.
Example 5b shows the graph derived from the chart in
Example 5a. It indicates that the intervals -8, -6, -4, -2, 1, and
+3, can be sequenced to compose a melody that will produce a
correct canon at the upper fifth. Note that +5 is not on the
graph because it was not involved in any valid cases. One
traces a path following the arrows from interval to interval,
never immediately taking the same two-way arrow in direct
succession unless it is marked with an asterisk. The adjacency
table in Example 5b tallies the number of different intervals to
and from each of interval on the graph. Some canons derived
from the graph are given in Example 6. The first of these
follows the path -8, 3, -2, 3, 3, -4, 3 on the graph. The last,
written by Robert Gauldin, is embellished and ends with a
typical 16th-century cadence.
The chart and graphs of Example 5 form a canon system
for N = +5. A change of N produces a canon system with dif-
ferent linear intervals. For example, the canon system for N =
-2 allows only the linear intervals of -4, -2, 2, 4, and 5.
6 Indeed, it would appear that the differences between modal and tonal
canons could be defined by what criteria are used to prune such charts. How-
ever, the cases pruned from Ex. 5a are invalid in both modal and tonal
polyphony. Of course, in a context not limited to first-species, two-voice
canon the differences between the two styles (including each's own incon-
sistencies) would need to be explicitly distinguished.
Example 6. Canons generated by the +5 canon
system.
Example 7. Canons generated by the -2 canon

sytem.
Integral 45
Example 7 presents two canons from this system. Note that the
latter has D = 2.
The table in Example 8 gives the melodic intervals for all
canon systems for N = -8 to +8. The column heads give the
values of N. The row heads provide the legal, that is, conso-
nant vertical intervals. Each column provides the melodic in-
tervals that will yield the vertical intervals of the row heads on
the left. This means that the content of the table position at the
intersection of row x and column y added to the content of
row head x will yield the content of column head y. Only
melodic intervals that are permitted in traditional counterpoint
are given; thus, ±7 and all intervals larger than 8 are omitted.
Looking at the column under N = +5, we find the melodic in-
tervals of the N = +5 canon system. However, one of this
column's intervals is not in the graph of the N = +5 canon
system, namely +5; it was ruled out in the chart in Example 5a
because it consistently produced invalid cases. Thus the
melodic intervals in the table may be pruned from a particular
canon system for stylistic reasons. Nevertheless, the table al-
lows us to compare the melodic intervals in various canon
systems. For instance, the intersection of the melodic intervals
for N = 3 and N = 5 produces the set of melodic intervals: {-8,
-6, -4, 1, 3, 5}. Thus, melodies that only use these intervals
may satisfy both canon system N = 3 and canon system N =
5. The table also indicates that canon systems with inverse Ns
have inverse melodic intervals, a point I will amplify later in
the discussion of canon groups.
In any case, Example 8 highlights the fact that each
canon system has its own unique melodic signature. This is
how a melody with a particular canonic potential can be im-
mediately recognized. For instance, as we have seen, melodies
that sequence intervals -8, -6, -4, -2, 1, and/or +3, will produce
a correct canon at N = +5, at the fifth above. By way of con-
trast, melodies that sequence intervals -4, -2, 2, 4, and/or 5 will
produce a correct canon at N = -2, the second below. In addi-
tion, the canon system graphs generate motivic signatures for
each system. For instance, for N = +5, the interval sequences
<3, -2> or <-2,-2> generate correct canons; for N = +2, <4,-2>
Si
I
Hi
C II
1
1
00
Integral 47
will generate a correct canon. So when we hear a note se-

quence of intervals <3,-2> we know it can be used in a subject
for a canon at the fifth; and when we hear a <4, -2> sequence,
we know it can be part of a subject for a canon at the second.
The canons of Examples 6c and 7b suggest paths for
further inquiry. First, what of canons with D > 1? Second, how
can one embellish first-species canons to make them more
musically serviceable? Third, how is dissonance treatment and
ornamentation accomplished within the graph methodology?
While I do not have enough space to address these questions
completely, it should be understood at the outset that they are
highly intertwined. The process of contrapuntal ornamenta-
tion or diminution in effect divides the beat so that a canon
originally at a delay of X beats, if suitably renotated, will be at
a delay of at least twice X beats.
If one would like to generate a first-species canon without
ornamentation where D is 2 or greater, there are two choices.
The first is necessary but not sufficient. One can interleave two
canons alternating them beat by beat to produce a canon with
D = 2. In such a canon, taking every other beat will yield a in-
dependent canon at D = 1. Example 7b was composed using
this method. Since beats in each of the two generating canons
are not actually adjacent in the resulting canon, some of the
cases originally pruned from the chart for stylistic reasons
might be reinstated in this environment. On the other hand,
care must be taken to make sure that the moves from one
canon stream to the other, between adjacent notes in the result-
ing canon, obey contrapuntal and melodic rules. I concede
that such contingencies make this method a bit dicey for gen-
erating tonal canons; but in the case of post-tonal canons,
where rules of voice leading and dissonance treatment are not
at issue, this method is a perfectly valid means of generating
canons for arbitrarily large values of D.
A better and wholly general method to generate
tonal/modal canons when D is greater than 1 is to generate
charts that sequence two or more intervals and list the con-
catenations of these sequences in the body of the chart. Once
again, cases that deviate from stylistic conventions are pruned.
48 Integral
While this will generate graphs, the number of cases to con-

sider grows very quickly so that resulting graphs become too
cumbersome for humans to use. Of course, adjacency matrices
and/or linked lists can substitute for graphs, but these data
structures are more appropriate in computational studies.
Fortunately, methods that generate canons with D > 1 in-
volving ornamentation and/or diminution are equally general
but less delicate or complex. Here one has two basic options.
One generates a canon from a graph, then embellishes it ac-
cording to standard methods such as inserting passing and
neighboring tones and the like. Such a two-pass method works
adequately enough, but involves backtracking and models
canon composition as two qualitatively different cognitive
stages. However, there is a more elegant solution. One con-
structs a canon system not only from the melodic intervals, but
from their transformations under standard ornamentation.
To illustrate this method, examine Example 9. In Ex-
ample 9a we have the interval 3 written in the context of the
canon system where D =1 and N = +5. We can transform the
case on the left into the case on the right by inserting a pass-
ing tone in each ascending third. We call the transformed 3
"3pt". Similarly, in Example 9b, the interval -2, is transformed
into a delayed "resolution" called -2s. Now we know from the
graph of the +5 canon system in Example 5b, that interval 3
can be preceded by the intervals -8, -6, -4, -2, and 1 and that 3
can move to the same set of intervals. We then test pairs of in-
tervals that begin or end with 3 (using the intervals just cited)
substituting 3pt for 3 to see if any stylistic constraints for
counterpoint or melody are violated. Since the test shows that
every succession is legal, we can always substitute 3pt for 3 in
the process of generating a canon from the graph. Thus, the
transformation of 3pt for 3 is context-free, that is, valid in all
cases of canon system N = +5. In the case of -2s, when we
substitute -2s for -2 in all legal interval successions involving
-2, not all cases yield legal contrapuntal textures. Example 9c
lists all the cases; intervals that precede -2 are written as pre-
fixes to -2s and their cases are illustrated on the left. Intervals
that can follow -2 are written as suffixes to -2s and their cases
Example 9. Embellished cases.
50 Integral
are written on the right. Most of the cases in Example 9c are

correct; only the two bottom cases on the right deviate from
ordinary practice. These are marked with an X and indicate
that the -2s transformation is context-sensitive. The first of
these is the sequence -2s, 3 and produces a 4-3 suspension,
which in many accounts of modal counterpoint is prohibited
in two-voice textures; however, it should be remembered that
two-voice 4-3 suspensions are tolerated in 18th-century
polyphony, so this case could be retained. The recursive -2s,
-2s case is incoherent since the first -2s makes it impossible
for the second -2s transformation to find a half-note to delay.
Nevertheless, I have written one way to implement the repeti-
tion of -2s which sufficiently disturbs the rhythmic stability of
the case for it to be omitted. Looking over the rest of the
cases, we see that we have implemented 7-6 and 6-5 progres-
sions with the -2s transformation.
Now, we can add the remaining cases involving -2s and
3pt to the graph of canon system +5. The result is in Example
10a. The graph is interpreted as before and automatically
generates correct canons with ornamentation. A canon gener-
ated from the graph is given in Example 10b.
Of course, we could have added 3pt and -2s as row and
column heads in the original chart for the +5 canon system in
Example 5a. The cases in Example 9 would then be cases in
the chart and the pruning of -2s followed by 3 or itself would
then be accomplished on the chart, rather than ex post facto as
above. Other transformations, such as lower neighbor notes or
runs of ornamental eighth-notes, k la third species, could be
added as well.
As a transition to considering non-tonal canons, Example
lla illustrates basic symmetries among first-species canons
and canon systems: how serial transformations affect a canon.
In this discussion and throughout the rest of this paper, I shall
use the term inversion only in its serial sense, to denote the
transformation of a pitch or pitch-class onto its inverse. The
exchange of voices, as in so-called "invertible counterpoint,"
will be called voice exchange.
Example 10
a. Graph for canon at fifth above with ornamentation.
I x / \ 7s* \ ^""^sV5*5
(3or3pt) <
/\
< \< j|c
\ = repeatable arrow
b. Canon based on graph in Example 10a.

Sequence from graph 3 -2s 1 3 3pt -2s -2 -2
a _ J I J J«- J J..
Example 11.
Integral 53
The canon fragment at the upper left is transformed

under R, I, and RI. The canon group, the simple mathematical
group that underlies such transformations is given in Example
lib. It indicates that: (1) every canon system for N is related
to that of -N by inverting the linear intervals in the former's
chart and graph; and that (2) under RI, the last four arguments
of a canon descriptor remain invariant. This means canons
related by RI are from the same canon system. In turn, this
implies that first-species canon systems are not temporally
teleological to within inversion. Lack of overall temporal
teleology is typical of a good deal of 15th- and 16th-century
polyphonic music; in contrast, Baroque music has such teleol-
ogy built in due to its chord grammars and embedded har-
monic/linear syntax. Of course, the introduction of ornamen-
tation into canons in any style renders them temporally non-
symmetric, for, after all, ornamentation delays and anticipates
otherwise temporally symmetric musical events. This implies
that transformational symmetry of the canon group of Ex-
ample lib does not invariably turn one correct ornamented
canon into another. Composers will see this outcome as an in-
stance of a general composition principle: the musical realiza-
tion of a symmetric compositional design usually breaks its
symmetry in the process of realization.
The extension of canon systems to non-tonal contexts is
accomplished in two stages. First, the limits of tonal counter-
point and harmony are removed so that any pitch interval may
be used as a melodic interval (in the canon subject) or as a
vertical interval. Nevertheless this extension does not engage
pitch-class relations as in serial music so that complementary
and compound interval equivalence does not hold.
In the sequel, pitches, pitch-classes and intervals will be
denoted as in most theories of atonal music, so that 0 denotes
a unison, +7 denotes an interval of seven semitones, and so
forth. We shall use t and e as for the pitch-classes 10 and 11,
respectively. I shall also discontinue the use of the term
"melodic interval" since the sequence of linear intervals in a
non-tonal canon subject need not form a "melody."
•• 1 s3i_ I ::e J J
s3i_ ::e *-ii en3
? -12- ^ =«- 3 as=-- $ =Ji---
+
* "S- -i*- ^3^. -3:..
SO
66 m l k I s ^ =
I „- *-■• 4- *■•• •? ..I;- ? I ];-
"TW "1 "1 ["Ji

« ~ -j> ii» - "■'» ♦ - """ ♦ - - f 1-
~ ° ---ik 3 J--,
£
a" ♦„. I- *-•• i *- ? T.!;-
■a •■> -1. -4
'E.
«s,j^ <u\l? <'*ll? tIwP

t « t ?
o V 2
I 1 11
Integral 55
Example 12b. Graph for pitch canon system

(-2, {-11,-6,6, 11}).
Examples 12a and 12b provide the chart and graph for a
canon system with D = 1 and N = -2 with vertical intervals -11,
-6, 6 and 1 1 . The comes will follow the dux at one beat two
semitones lower; the vertical intervals will either 6 or 1 1 semi-
tones wide. We call such systems pitch canon systems. Differ-
ent pitch canon systems are denoted by the couple (N, I),
where I is the set of ordered vertical intervals. The pitch canon
system at hand is (-2, {-11, -6, 6, 11}). As in any canon sys-
tem, the vertical intervals in set I, when subtracted from N,
produce the set of linear intervals in the system. So the linear
intervals are: +9 = N -(-11) or -2 -(-11); +4 = -2 -(-6); -8 =
-2 -(6); and -13 = -2 -(11). The chart of this pitch canon
system has the four linear intervals as row heads. As in tonal
systems, certain successions of intervals on the chart may be
ruled out; yet here such cases are pruned from the chart by
composer fiat rather than tradition. In this case, proximate
octaves and "uninteresting" sonorities and contours (for in-
stance, whole-tone scales) are not permitted. Some canons de-
rived from this system are given in Example 13, the last of
which is elaborated into a more typical 20th-century surface.
Note that the musical realization in Example 13c2 camou-
flages the canonic voices of Example 13cl in order to articu-
late various sets of adjacent canon pitches as instances of
Example 13. Canons at -2 below.
Vertical intervals = {-11, -6, 6, 11}.
Integral 57
(pitch-class) set-classes 3-3[014] at the beginning and 3-

8 [026] at the end.
The second stage of non-tonal extension involves the
mapping of pitch to pc. Hence, the charts are no longer writ-
ten on the staff but with pc integers. These systems are called
pc canon systems. The voices in the canons they generate are
properly called lynes. (A lyne is an uninterpreted string of
pitch-classes. For a lyne to be heard as a "voice," its pcs must
be interpreted in some coherent global way, such as in the
same register, in a given instrument, at a certain loudness, and
so forth.) Example 14 provides a chart, graph, and pc canon
for a pc canon system. Here the linear intervals are ordered pc
intervals. The reader will note, that unlike in pitch canon sys-
tems, the vertical intervals (set I) are unordered intervals. I
shall use the standard term ic (for interval-class) to stand for
unordered pc intervals. (One could continue to use ordered
intervals to model voice crossing, but I do not, if only because
register is undefined among pitch classes.) The linear ordered
intervals are derived by subtracting each member of each ver-
tical ic from N.
While a pc canon system can be constructed directly as
described above, a system can be derived from a pitch canon
system as well. The pc canon system of Example 14 is gener-
ated from the pitch canon system in Examples 12 and 13
simply by taking the former system mod 12: N changes from
-2 to t and the intervals -11, -6, 6, 11 are respectively mapped
to the intervals 1, 6, 6, e, members of icl and ic6. Note that the
set of vertical intervals has only three distinct members since
the pitch intervals -6 and +6 both map to pc interval 6. This is
the reason the chart for this pc system measures 3 by 3,
instead of 4 by 4.
Certain cases in the chart of Example 14a are pruned ac-
cording to roughly the same criteria as before; only hexa-
chords of the set-class (SC) 6-2[012346] are chosen. Thus the
resulting canons will be saturated by this hexachordal sonor-
ity. The graph is symmetric, allowing any sequence without
immediate repetition of the linear pc intervals 4, 9, or e.
Example 14.
a. Pc canon system (t, {icl, ic6}).

(Pc version of pitch canon system (-2, {6,11}) in Ex. 12a.)
e, eD e~4 eT5
e: t 9 t 9 8 t 9 l t 9 6
Oe 0 e t 0e3 0e8
4, e 4, 4 U 4, 9
4: t 2 t21 t 2 6 t2e
04 043 0 4 S 041
9, e 9, 4 9, 9 S
9: t7 t76 t7e t74
09 098 091 0 9 6
I SC(4-10) SC{
LEGEND
D = duplicated pcs.
U = "uninteresting" SC.
S = singular hexachordal SC.
b. Graph for pc canon system (t, {icl}, {ic6}).
e adjacency table
jf w Ic Sto #from
aC
c. A canon generated by the pc canon
Intervals: e 9 4 e 4 e
Dux: 0 e 8 0 e 3 2
Comes: t 9 6 t 9 10
Vert, ics 116 16 1
Integral 59
The symmetry of any canon system graph is guaranteed

so long as the composer does not prune any cases from the
system's chart, so that all linear intervals are context free. After
pruning, however, the intervals on the graph are context sensi-
tive. Graph symmetry ensues if the composer prunes sets of
cases that are symmetric around the main diagonal of the
chart; that is, the case in the xth row and yth column must be
pruned if and only if the case in the yth row and xth column
is pruned. The symmetry of the present graph results from
deleting cases along the main diagonal of the chart in
Example 14a.
Pc canon systems have different properties than pitch
canon systems. Among these is the underlying group structure
relating pc canons. While registral interchange (i.e. double
counterpoint) does not in general guarantee correct contra-
puntal relations between voice pairs in tonal and modal
polyphony (especially when the voices are in canon), in pc
canon systems ics do not change when the "bottom" and "top"
lynes are exchanged. We therefore compose a pc canon group
of four transformations: To, RI, voice exchange, and RI with
voice exchange. The pc canon group preserves N. This con-
trasts with the canon group of Example lib which includes R
and I as transformations and does not preserve N. Four canons
generated from a canon from the pc canon system (4, {icl,
ic6}) interrelated by the pc canon group are given in Example
15a. The exchange operation allows the linkage of the four
canons to form canonic fallowings. These are two concate-
nated canons, linked so that the last note of the first canon's
comes voice is played with the first note of the second canon's
dux. We constrain the link so it preserves the vertical ics of its
generating canons. However, canon linkage may not preserve
N. As an example see Example 15b which links canons related
by Tol; note that the canon interval (N = 4) changes to its in-
verse (8) after the link. To get canon followings that preserve
N, one can enlarge the set of four canons related by the pc
canon group to 48 under pc transposition and choose canons
that utilize a link interval from the vertical intervals of the pc
Example 15.
a. hour pc canons related by the pc canon group.
(no exchange) with exchange

N = 4 N = 4
P: 0 3 6 4 9 = dux T4P: 4 7 t 8 1 = old comes
T4P: 4 7 t 8 1 = comes p: 0 3 6 4 9 = old dux
N = 4 N = 4
RT0IP: 3 8 6 9 0 = old dux RTglP: e 4 2 5 8 = old comes
RTglP: e 4 2 5 8 = old comes RTqIP: 3 8 6 9 0 = old dux
b. A canonic following of two canons related by TqI.
P: T0IP:
0 3 6 4 9 0 9 6 8 3
4 7 t 8 1 8 5 2 4 e link ic = 1 (at T)
T4P: t T8IP:
Note: the canon interval N changes from 4 to 8 after the link.
c. Canonic followings that preserve N = 4.
P: T7P:
0 3 6 4 9 7 t 1 e 4
4 7 t 8 1 a 2 5 3 8 link ic = 6 (at t)
T4P: t TeP:
P: RT9IP:
0 3 6 4 9 0 5 3 6 9
4 7 t 8 1 4 9 7 t 1 link ic = 1 (at T)
T4P: T RTilP:
P: RT2P:
0 3 6 4 9 2 5 8 6 e
4 7 t 8 1 6 9 0 t 3 link ic = 1 (at T)
T4P: T RT6P:
P: RT4IP:
0 3 6 4 9 7 0 t 1 4
4 7 t 8 1 e 4 2 5 8 link ic = 6 (at T)
T4P: T RTglP
Integral 61
canon system. Some canonic fallowings that preserve N are

given in Example 15c.
Example 16. Number of set classes that can comprise the pc

content of pc canons.
# of set-classes available # of set-classes

n m for two- voice canons. available for three-or-
3 I 12 I 5 I 5
4 TJT 13 13
T~^1T 15 " 15
"Tj 5IT 30 33 ~ ~
~1 W 30 31 ~
~g W* 25 29 ~
9 I 12 I 11 | 12
n: Set class cardinality.
m: Number of set classes containing p
Since, allowing for pc repetition, th

12 ordered intervals, there are 1,524
tems.7 This contrasts with the 14 or so
tems and the thousands of pitch ca
Since we have used set-classes formed
cent pcs in corresponding order po
comes as a criterion for pruning certai
pc canon system, it is interesting that
be found adjacently in any canon from
any values of D and N in any number
from the fact that a pc canon's content
combination" of the content of its dux voice and the union of
pcs that begin each voice.8 Since some set-classes cannot be
7 1,524 = (27 - 1) X 12. For 3 ics there are 21 X 12 = 252 pc canon sys-
tems; C(7,3) = 21.
8 See Richard Cohn, Transpositional Combination in Twentieth-Century
Music (Eastman School of Music, University of Rochester: Ph.D. diss.,
1986). The appendices in Volume 2 contain useful tables of primes, idempo-
tent SCs, and the like.
62 Integral
generated by transpositional combination - those designated

"primes" and "semi-primes" by Richard Cohn - they cannot
occur in any pc canon as adjacent pcsets. A table indicating
the number of set-classes that can comprise the pc content of
canons of two voices or more is found in Example 16. The
table shows, for instance, that of the 50 hexachordal set-classes
only 30 can be produced by the pcs of a two- voice pc canon.
Further, Cohn's "idempotent" set-classes allow certain pc
canons to exhaust only a subset of the aggregate - regardless
of the number of voices - providing one confines their sub-
jects to certain set-classes and uses only certain values for N.
Example 17. Fragments from a canon by inversion at the fifth

above.
Up to now, we have only considered canons by transposi-

tion. What about canons by inversion? These canons have
their dux followed by the comes transformed under inversion
and transposition. Unfortunately, there seems to be no way to
generate inversional canons with the methodology developed
herein. The transpositional canon systems considered so far
are based on the condition that the melodic intervals in the
dux and comes are mapped one to one to the vertical intervals
between them. This is implied by the fact that N equals the
melodic interval j plus the vertical interval i, as in Example 4a.
Now consider the three cases in Example 17. These are canon
fragments from a hypothetical canon by inversion at the fifth
above at a delay of one beat. The first two cases show that the
melodic interval is associated with two different vertical inter-
vals while the first and last cases show that the vertical interval
is associated with two different melodic intervals. Thus there is
Integral 63
no mapping at all from the melodic intervals to the vertical

intervals or vice versa. This means that a given melodic inter-
val generates a particular vertical interval only if we know the
first pitch of the interval. So it is impossible to generate a
graph of interval successions that can be traced to generate all
and only subjects for inversional canons. As a result, canon
systems can be only tangentially related to George Perle's
Twelve-Tone Tonality? whose "cyclic sets" are canons related
under rotation, retrograde, transposition and inversion opera-
tions.
Example 18.
a. A minimal inversional canon fragment.
f , N = canon interval,
comes:' Jc , a, b = notes in dux.
* I c, d = notes in comes.
V -j J j = melodic interval from a to b
v - -j = melodic interval from c to d.
b f i = sum of b + c.
dux: ifj N = M ifj>0
J/n N = i-(j+2)ifj<-l
b. Canon by inversion; N=6.
Qy tf ° tt ° " o o
^ <» o o o ^ " »
j = melodic intervals dissonance
+2 +2 -2 -2 +4 1
i = harmonic sum
7 7 5 5 9 6
N = i + (-j) = 6
^See George Perle, Twelve-Tone T

fornia Press, 1977).
64 Integral
Nevertheless, in the interest of theoretic closure, we can

model canons by inversion using sums of pitches. Example
18a analyses a canon fragment from a tonal/modal canon by
inversion at interval N. Pitches a to b form the interval j, and
pitches c to d form the inverse interval -j. The variable i is the
sum of the simultaneous notes b + c. Finally, N = i - j. Exam-
ple 18b provides a canon by inversion at the sixth. As the ex-
ample shows, N = 6 and the sums of the simultaneous pitches
minus the preceding interval in the dux equals N or 6. The in-
correct - that is, dissonant - last simultaneity helps assure us
that the model works for all sums and intervals, dissonant or
not.
The formal study of canon systems and their visual repre-

sentations as transition graphs conjures up associations with
David Lewin's generalized musical intervals and transforma-
tions.10 Certainly, the set of ordered pitch or pitch-class inter-
vals and their concatenations forming intervals from the same
set under addition which provides the context for a canon
system is the very GIS which has motivated Lewin's general-
izations. Were the canon systems themselves instances of a
generalized interval system, they would be matched by a pow-
erful analytic methodology. Unfortunately, however, this is
not the case for two reasons. First, the linear and vertical inter-
vals of a canon system are not entities related by generalized
intervals or transformations, but only by possible succession.
Second, the intervals of the canon system graph nodes are not
"things" as in Lewin's transformation networks, but simply
equivalence classes of ordered pitch or pc pairs. Two pairs
<a,b> and <c,d> are equivalent if b-a = d-c. This equivalence
is coextensive with the concept of ordered pitch interval. Of
course, the canon groups of Examples lib and 15a can be
modeled by a GIS and extended to model the invariances of
invertible counterpoint for any number of voices.
In an important sense, we have been studying the rela-
tions between melodic structure and polyphonic opportunity.
*°See David Lewin, Generalized Musical Intervals and Transformations

(New Haven: Yale University Press, 1987).
Example 19. An enigmatic canon.
a. Canon a 4: dedie A Monsieur Houdemann et compost par

J. S. Bach.
l||g||5iin / 1 y [ f f ^J^TVr t- hop ■ r ||5^r P^y
b. Solution to the above.
■
*
7fLiii,.
-h Lr -h Lr- ^ZZI
^ZZI 1 , 11II*
1 ,, iI , ,iIIi .II, .I ,- iT m i
I I
7 <bV]r -h Lr ^ZZI I - J , 1 r II* , I J I I I IJ II . J , I - I
v;^fJiJf^ ■ r r Ir tt$ ^ ^ ^
(After TAc Bach Reader, edited by Hans T. David and Arthur
Mendel, pp.111, 402.)
66 Integral
With more experience with tonal canon systems, we could de-

velop the ability to perform musical parlor tricks to astound
our colleagues, such as deftly transforming the enigma canon
in Example 19a into its solution in Example 19b.11 But there
is more at stake here, for canon systems and other similar
types of musical dependencies give us insight into the cogni-
tion of master musicians who accomplish complex musical
tasks effortlessly in the real time of musical improvisation or
listening. Specifically, the canon systems help to explain how
J. S. Bach was able to perform the contrapuntal predictions
described by Carl Philipp Emanuel Bach in a letter of 1774 to
Johann Nicolaus Forkel:
When he listened to a rich and many-voiced fugue,

he could soon say, after the first entries of the sub-
ject, what contrapuntal devices it would be possible
to apply, and which of them the composer by rights
ought to apply, and on such occasions, when I was
standing next to him, and he had voiced his surmises
to me, he would joyfully nudge me when his
expectations were fulfilled.12'13
11 The canon is transcribed into modern notation from Mattheson's Der

vollkommene Capellmeister of 1739. The solution in Example 19b is given
in The Bach Reader, revised edition, Hans T. David and Arthur Mendel, eds.
(New York: Norton, 1966), p. 402.
nThe Bach Reader, p. 277.
13 I would like to thank Daniel Harrison and Dora Hanninen for reading
early drafts of this paper and offering suggestions for improvement. My
thanks also go to Norman Carey who added a few fine touches and
corrections.

The Structure of First-Species Canon in Modal - Tonal An Atonal MusicsMorris

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

The Structure of First-Species Canon in Modal - Tonal An Atonal MusicsMorris

Загружено:

Авторское право:

Доступные форматы

The Structure of First-Species Canon in Modal,

Tonal and Atonal Musics

It has been frequently and appropriately noted that serial

*See Serge Taneiev, Convertible Counterpoint (Moscow, 1909).

abstract canons at various intervals, inversions, retrogrades, and

Example 2. Canon at the fourth below and at a delay

3 See Robert Morris, "Generalizing Rotational Arrays," Journal of Music

Consider Example 2. It contains a note-agairist-note

where P is the name of the canon tune or subject, N is the

Example 3: Canon descriptors.

General canon descriptor: (P, N, D, I, J)

D = time delay (in beats)

Descriptor for canon in Example 2:

In modal and tonal music, the vertical intervals allowable

are specified in tonal music, they control the values of j for a

Example 4a. A minimal canon fragment.

Example 4b. Some values for j and i that generate

To generate canons longer than that of Example 4b, we

<u ^-> cd >

linear interval succession -4, 3 that generates a three-note

Example 5b. Graph for canon at fifth above.

Of course, not all cas

Example 7. Canons generated by the -2 canon

will generate a correct canon. So when we hear a note se-

While this will generate graphs, the number of cases to con-

are written on the right. Most of the cases in Example 9c are

a. Graph for canon at fifth above with ornamentation.

b. Canon based on graph in Example 10a.

The canon fragment at the upper left is transformed

? -12- ^ =«- 3 as=-- $ =Ji---

I „- *-■• 4- *■•• •? ..I;- ? I ];-

"TW "1 "1 ["Ji

«s,j^ <u\l? <'*ll? tIwP

Example 12b. Graph for pitch canon system

(pitch-class) set-classes 3-3[014] at the beginning and 3-

a. Pc canon system (t, {icl, ic6}).

b. Graph for pc canon system (t, {icl}, {ic6}).

Vert, ics 116 16 1

The symmetry of any canon system graph is guaranteed

(no exchange) with exchange

b. A canonic following of two canons related by TqI.

Note: the canon interval N changes from 4 to 8 after the link.

c. Canonic followings that preserve N = 4.

canon system. Some canonic fallowings that preserve N are

Example 16. Number of set classes that can comprise the pc

# of set-classes available # of set-classes

Since, allowing for pc repetition, th

generated by transpositional combination - those designated

Example 17. Fragments from a canon by inversion at the fifth

Up to now, we have only considered canons by transposi-

no mapping at all from the melodic intervals to the vertical

b. Canon by inversion; N=6.

^See George Perle, Twelve-Tone T

Nevertheless, in the interest of theoretic closure, we can

The formal study of canon systems and their visual repre-

*°See David Lewin, Generalized Musical Intervals and Transformations

a. Canon a 4: dedie A Monsieur Houdemann et compost par

l||g||5iin / 1 y [ f f ^J^TVr t- hop ■ r ||5^r P^y

b. Solution to the above.

7 <bV]r -h Lr ^ZZI I - J , 1 r II* , I J I I I IJ II . J , I - I

With more experience with tonal canon systems, we could de-

When he listened to a rich and many-voiced fugue,

11 The canon is transcribed into modern notation from Mattheson's Der

I „- -■• 4- ■•• •? ..I;- ? I ];-